Max Robotics — Global Service Robot Directory + US Certification

About

Nur Muhammad Mahi Shafiullah is a robotics and machine learning researcher whose work sits at the intersection of imitation learning, visual representation, and deployable robot systems. His research addresses a central challenge in modern robotics: enabling robots to learn generalizable behaviors from limited, imperfect, or uncurated demonstrations without requiring task-specific engineering. Shafiullah's most cited work, "The Surprising Effectiveness of Representation Learning for Visual Imitation" (82 citations), demonstrated that strong visual representations can dramatically simplify the complexity typically required for visual imitation learning. His development of CLIP-Fields (68 citations) introduced a weakly supervised implicit scene model leveraging vision-language embeddings for robotic memory, enabling robots to perform semantic search and spatial reasoning. Through Behavior Transformers (31 citations), he tackled the multimodal nature of human demonstration data using transformer architectures capable of capturing diverse behavioral modes. More recently, his OK-Robot project (41 citations) showcased how integrating open-knowledge vision and language models can enable zero-shot object manipulation in real homes — a meaningful step toward practical generalist robots. His broader research agenda, reflected in works like "On Bringing Robots Home" and "Robot Utility Models," consistently pushes toward robots that are deployable, adaptable, and genuinely useful in unstructured everyday environments.